| The Ethernet Evolution   The 180 Degree Turn   (C) Herbert Haas 2005/03/11   1 Download from Www.Somanuals.com. All Manuals Search And Download.   History: Initial Idea   ꢀ Shared media ꢁ CSMA/CD as access algorithm   ꢀ COAX Cables   ꢀ Half duplex communication   ꢀ Low latency ꢁ No networking nodes   (except repeaters)   ꢀ One collision domain and also one broadcast domain   10 Mbit/s shared   by 5 hosts ꢁ 2   Mbit/s each !!!   3 (C) Herbert Haas 2005/03/11   The initial idea of Ethernet was completely different than what is used today   under the term "Ethernet". The original new concept of Ethernet was the use of   a shared media and an Aloha based access algorithm, called Carrier Sense   Multiple Access with Collision Detection (CSMA/CD). Coaxial cables were   used as shared medium, allowing a simple coupling of station to bus-like   topology.   Coax-cables were used in baseband mode, thus allowing only unicast   transmissions. Therefore, CSMA/CD was used to let Ethernet operate under the   events of frequent collisions.   Another important point: No intermediate network devices should be used in   order to keep latency as small as possible. Soon repeaters were invented to be the   only exception for a while.   An Ethernet segment is a coax cable, probably extended by repeaters. The   segment constitutes one collision domain (only one station may send at the same   time) and one broadcast domain (any station receives the current frame sent).   Therefore, the total bandwidth is shared by the number of devices attached to the   segment. For example 10 devices attached means that each device can send 1   Mbit/s of data on average.   Ethernet technologies at that time (1975-80s): 10Base2 and 10Base5   3 Download from Www.Somanuals.com. All Manuals Search And Download.   History: Multiport Repeaters   ꢀ Demand for structured cabling (voice-grade   twisted-pair)   ꢂ 10BaseT (Cat3, Cat4, ...)   ꢀ Multiport repeater ("Hub") created   ꢀ Still one collision domain   ("CSMA/CD in a box")   4 (C) Herbert Haas 2005/03/11   Later, Ethernet devices supporting structured cabling were created in order to   reuse the voice-grade twisted-pair cables already installed in buildings. 10BaseT   had been specified to support Cat3 cables (voice grade) or better, for example   Cat4 (and today Cat5, Cat6, and Cat7).   Hub devices were necessary to interconnect several stations. These hub devices   were basically multi-port repeaters, simulating the half-duplex coax-cable, which   is known as "CSMA/CD in a box". Logically, nothing has changed, we have still   one single collision and broadcast domain.   Note that the Ethernet topology became star-shaped.   4 Download from Www.Somanuals.com. All Manuals Search And Download.   History: Bridges   ꢀ Store and forwarding according destination MAC   address   ꢀ Separated collision domains   ꢀ Improved network performance   ꢀ Still one broadcast domain   Three collision   domains in this   example !   5 (C) Herbert Haas 2005/03/11   Bridges were invented for performance reasons. It seemed to be impractical that   each additional station reduces the average per-station bandwidth by 1/n. On the   other hand the benefit of sharing a medium for communication should be still   maintained (which was expressed by Metcalfe's law).   Bridges are store and forwarding devices (introducing significant delay) that can   filter traffic based on the destination MAC addresses to avoid unnecessary   flooding of frames to certain segments. Thus, bridges segment the LAN into   several collision domains. Broadcasts are still forwarded to allow layer 3   connectivity (ARP etc), so the bridged network is still a single broadcast domain.   5 Download from Www.Somanuals.com. All Manuals Search And Download.   History: Switches   ꢀ Switch = Multiport Bridges with HW acceleration   ꢀ Full duplex ꢁ Collision-free Ethernet ꢁ No CSMA/CD   necessary anymore   ꢀ Different data rates at the same time supported   ꢂ Autonegotiation   ꢀ VLAN splits LAN into several broadcast domains   Collision-free   plug & play   1000 Mbit/s   scalable Ethernet !   100 Mbit/s   100 Mbit/s   10 Mbit/s   6 (C) Herbert Haas 2005/03/11   Several vendors built advanced bridges, which are partly or fully implemented in   hardware. The introduced latency could be dramatically lowered and   furthermore other features were introduced, for example full duplex   communication on twisted pair cables, different frame rates on different ports,   special forwarding techniques (e,g, cut through or fragment free), Content   Addressable Memory (CAM) tables, and much more. Of course marketing rules   demand for another designation for this machine: the switch was born.   Suddenly, a collision free plug and play Ethernet was available. Simply use   twisted pair cabling only and enable autonegotiation to automatically determine   the line speed on each port (of course manual configurations would also do). This   way, switched Ethernet become very scalable.   Furthermore, Virtual LANs (VLANs) were invented to split the LAN into several   broadcast domains. VLANs improve security, utilization, and allows for logical   borders between workgroups.   6 Download from Www.Somanuals.com. All Manuals Search And Download.   Today   ꢀ No collisions ꢁ no distance limitations !   ꢀ Gigabit Ethernet becomes WAN   technology !   ꢂ Over 100 km link span already   ꢀ Combine several links to "Etherchannels"   ꢂ Acts as single link from the spanning-tree view   • Cisco: Port Aggregation Protocol (PAgP)   • IEEE 802.1ad: Link Aggregation Control Protocol   (LACP)   1 Gbit/s or even 10 Gbit/s long reach connection !!!   7 (C) Herbert Haas 2005/03/11   Today, Gigabit and even 10 Gigabit Ethernet is available. Only twisted pair and   more and more fiber cables are used between switches, allowing full duplex   collision-free connections. Since collisions cannot occur anymore, there is no   need for a collision window anymore! From this it follows, that there is virtually   no distance limit between each two Ethernet devices.   Recent experiments demonstrated the interconnection of two Ethernet Switches   over a span of more than 100 km! Thus Ethernet became a WAN technology!   Today, many carriers use Ethernet instead of ATM/SONET/SDH or other rather   expensive technologies. GE and 10GE is relatively cheap and much simpler to   deploy. Furthermore it easily integrates into existing low-rate Ethernet   environments, allowing a homogeneous interconnection between multiple   Ethernet LAN sites. Basically, the deployment is plug and play.   If the link speed is still too slow, so-called "Etherchannels" can be configured   between each two switches by combining several ports to one logical connection.   Note that it is not possible to deploy parallel connections between two switches   without an Etherchannel configuration because the Spanning Tree Protocol (STP)   would cut off all redundant links.   Depending on the vendor, up to eight ports can be combined to constitute one   "Etherchannel".   7 Download from Www.Somanuals.com. All Manuals Search And Download.   What About Gigabit Hubs?   ꢀ Would limit network diameter to 20-   25 meters (Gigabit Ethernet)   ꢀ Solutions   ꢂ Frame Bursting   ꢂ Carrier Extension   ꢀ No GE-Hubs available on the market   today ꢁ forget it!   ꢀ No CSMA/CD defined for 10GE (!)   8 (C) Herbert Haas 2005/03/11   Remember: Hubs simulate a half-duplex coaxial cable inside, hence limiting the   total network diameter. For Gigabit Ethernet this limitation would be about 25   meters, which is rather impracticable for professional usage. Although some   countermeasures had been specified in the standard, such as frame bursting and   carrier extension, no vendor developed an GE hub as for today. Thus: Forget GE   Hubs!   The 10 GE specification does neither consider copper connections nor hubs. 10   GE can only run over fiber.   At this point please remember the initial idea in the mid 1970s: Bus, CSMA/CD,   short distances, no network nodes.   Today: Structured cabling (point-to-point or star), never CSMA/CD, WAN   capabilities, sophisticated switching devices in between.   8 Download from Www.Somanuals.com. All Manuals Search And Download.   MAC Control Frames   ꢀ Additional functionality easily integrated   ꢀ Currently only Pause-Frame supported   Always 64 bytes   8 bytes   6 6 2 2 44   4 preamble DA   SA   8808h   MAC-ctrl opcode   MAC-ctrl parameters FCS   MAC-ctrl opcode ........... Defines function of control frame   MAC-ctrl parameters .... control parameter data (always filled up to 44 bytes)   9 (C) Herbert Haas 2005/03/11   Different data rates between switches (and different performance levels) often   lead to congestion conditions, full buffers, and frame drops. Traditional Ethernet   flow control was only supported on half-duplex links by enforcing collisions to   occur and hereby triggering the truncated exponential backoff algorithm. Just let   a collision occur and the aggressive sender will be silent for a while.   A much finer method is to send some dummy frames just before the backoff   timer allows sending. This way the other station never comes to send again.   Both methods are considered as ugly and only work on half duplex lines.   Therefore the MAC Control frames were specified, allowing for active flow   control. Now the receiver sends this special frame, notifying the sender to be   silent for N slot times.   The MAC Control frame originates in a new Ethernet layer—the MAC Control   Layer—and will support also other functionalities, but currently only the "Pause"   frame has been specified.   9 Download from Www.Somanuals.com. All Manuals Search And Download.   Auto Negotiation   ꢀ Enables each two Ethernet devices to   exchange information about their   capabilities   ꢂ Signal rate, CSMA/CD, half- or full-duplex   ꢀ Using Link-Integrity-Test-Pulse-Sequence   ꢂ Normal-Link-Pulse (NLP) technique is used   in 10BaseT to check the link state (green LED)   ꢂ 10 Mbit/s LAN devices send every 16.8 ms a   100ns lasting NLP, no signal on the wire   means disconnected   10   (C) Herbert Haas 2005/03/11   Several Ethernet operating modes had been defined, which are incompatible to   each other, including different data rates (10, 100, 1000 Mbit/s), half or full   duplex operation, MAC control frames capabilities, etc.   Original Ethernet utilized so-called Normal Link Pulses (NLPs) to verify layer 2   connectivity. NLPs are single pulses which must be received periodically   between regular frames. If NLPs are received, the green LED on the NIC is   turned on.   Newer Ethernet cards realize auto negotiation by sending a sequence of NLPs,   which is called a Fast Link Pulse (FLP) sequence.   10   Download from Www.Somanuals.com. All Manuals Search And Download.   Fast Link Pulses   ꢀ Modern Ethernet NICs send bursts of   Fast-Link-Pulses (FLP) consisting of   17-33 NLPs for Autonegotiation   signalling   ꢀ Each representing a 16 bit word   ꢂ GE sends several "pages"   11   (C) Herbert Haas 2005/03/11   A series of FLPs constitute an autonegotiation frame. The whole frame consists   of 33 timeslots, where each odd numbered timeslot consists of a real NLP and   each even timeslot is either a NLP or empty, representing 1 or 0. Thus, each FLP   sequence consists of a 16 bit word.   Note that GE Ethernet sends several such "pages".   11   Download from Www.Somanuals.com. All Manuals Search And Download.   100 Mbit Ethernet Overview   IEEE 802.3u   IEEE 802.12   Signaling Schemes   Demand Priority   Fast Ethernet   100BaseX   Signaling   Fast Ethernet   100Base4T+   Signaling   100VG-AnyLAN   100BaseT4   (half duplex)   100BaseFX   100BaseTX   HP and AT&T   invention for real time   applications   "100BaseT"   12   (C) Herbert Haas 2005/03/11   The diagram above gives an overview of 100 Mbit/s Ethernet technologies,   which are differentiated into IEEE 802.3u and IEEE 802.12 standards. The IEEE   802.3u defines the widely used Fast Ethernet variants, most importantly those   utilizing the 100BaseX signaling scheme. The 100BaseX signaling consists of   several details, but basically it utilizes 4B5B block coding over only two pairs of   regular Cat 5 twisted pair cables or two strand 50/125 or 62.5/125-µm multimode   fiber-optic cables.   100Base4T+ signaling has been specified to support 100 Mbit/s over Cat3 cables.   This mode allows half duplex operation only and uses a 8B6T code over 4 pairs   of wires; one pair for collision detection, three pairs for data transmission. One   unidirectional pair is used for sending only and two bi-directional pairs for both   sending and receiving.   The 100VG-AnyLAN technology had been created by HP and AT&T in 1992 to   support deterministic medium access for realtime applications. This technology   was standardized by the IEEE 802.12 working group. The access method is   called "demand priority". 100VG-AnyLAN supports voice grade cables (VG) but   requires special hub hardware. The 802.12 working group is no longer active.   12   Download from Www.Somanuals.com. All Manuals Search And Download.   4B/5B Coding   MII   16 code   groups   0 0 0 1 4 x 25   Mbit/s   PCS   32 code   groups   4B/5B Encoder/Decoder   0 1 0 0 1 PMA   125 MBaud   13   (C) Herbert Haas 2005/03/11   The diagram above shows the basic principle of the 4B5B block coding principle,   which is used by 802.3u and also by FDDI. The basic idea is to transform any   arbitrary 4 bit word into a (relatively) balanced 5 bit word. This is done by a fast   table lookup.   Balancing the code has many advantages: better bandwidth utilization, better   laser efficiency (constant temperature), better bit-synchronization (PLL), etc.   Note that the signaling overhead is 5/4 ꢁ 12.5 %.   13   Download from Www.Somanuals.com. All Manuals Search And Download.   Gigabit Ethernet   Media Access Control (MAC)   Gigabit Media Independent Interface (GMII)   1000Base-X   1000Base-T   8B/10B encoder/decoder   encoder/decoder   1000Base-CX   Shielded   1000Base-LX   LWL   Fiber Optic   1000Base-SX   SWL   Fiber Optic   1000Base-T   UTP   Balanced   Copper   Cat 5e   IEEE 802.3ab   physical layer   IEEE 802.3z physical layer   14   (C) Herbert Haas 2005/03/11   Gigabit Ethernet has been defined in March 1996 by the working group IEEE   802.3z. The GMII represents a abstract interface between the common Ethernet   layer 2 and different signaling layers below. Two important signaling techniques   had been defines: The standard 802.3z defines 1000Base-X signaling which uses   8B10B block coding and the 802.3ab standard uses 1000Base-T signaling. The   latter is only used over twisted pair cables (UTP Cat 5 or better), while   1000BaseX is only used over fiber, with one exception, the twinax cable   (1000BaseCX), which is basically a shielded twisted pair cable.   BTW: The "X" stands for block coding.   14   Download from Www.Somanuals.com. All Manuals Search And Download.   GE Signaling   IEEE 802.3   Ethernet   IEEE 802.3z   Gigabit Ethernet   ANSI X3T11   Fibre Channel   FC-4   802.2 LLC   IEEE 802.2 LLC   upper layer mapping   CSMA/CD   or full duplex MAC   FC-3   common services   802.3 CSMA/CD   Reconciliation Sublayer   FC-2   signalling   802.3 PHY   PCS   FC-1   encoder/decoder   PHY   PMA   PMD   FC-0   interface and media   15   (C) Herbert Haas 2005/03/11   Gigabit Ethernet layers have been defined by adaptation of the LLC and MAC   layers of classical Ethernet and the physical layers of the ANSI Fiber Channel   technology. A so-called reconciliation layer is used in between for seamless   interoperation. The physical layer of the Fiber Channel technology uses 8B10B   block coding.   15   Download from Www.Somanuals.com. All Manuals Search And Download.   GE 8B/10B Coding   GMII   Only used   by   1 1 1 1 1 1 1 1 256 code groups   8 x 125 Mbit/s   1000BaseX   PCS   1024 code groups   8B/10B Encoder/Decoder   125 million code   groups per   second   1 1 1 1 1 1 1 1 1 1 PMA   1250 Mbaud   16   (C) Herbert Haas 2005/03/11   8B10B block coding is very similar to 4B5B block coding but allows fully   balanced 10-bit codewords. Actually, there are not enough balanced 10-bit   codewords available. Note that there are 256 8-bit codewords which need to be   mapped on 1024 10-bit codewords. But instead of using a fully balanced 10-bit   codeword for each 8-bit codeword, some 8-bit codewords are represented by two   10-bit codewords, which are sent in an alternating manner. That is, both   associated 10-bit words are bit-complementary.   Again, the signaling overhead is 12.5%, that is 1250 Mbaud is necessary to   transmit a bit stream of 1000 Mbit/s.   16   Download from Www.Somanuals.com. All Manuals Search And Download.   1000BaseX   ꢀ Two different wavelengths supported   ꢀ Full duplex only   ꢂ 1000Base-SX: short wave, 850 nm MMF   ꢂ 1000Base-LX: long wave, 1300 nm MMF or SMF   ꢀ 1000Base-CX:   ꢂ Twinax Cable (high quality 150 Ohm balanced   shielded copper cable)   ꢂ About 25 m distance limit, DB-9 or the newer   HSSDC connector   17   (C) Herbert Haas 2005/03/11   Gigabit Ethernet can be transmitted over various types of fiber. Currently (at   least) two types are specified, short and long wave transmissions, using 850 nm   and 1300 nm respectively. The long wave can be used with both single mode   (SMF) and multimode fibers (MMF). Only SMF can be used for WAN   transmissions because of the much lower dispersion effects.   Note that there are several other implementations offered by different vendors,   such as using very long wavelengths at 1550 nm together with DWDM   configurations.   The twinax cable is basically a shielded twisted pair cable.   17   Download from Www.Somanuals.com. All Manuals Search And Download.   1000BaseT   ꢀ Defined by 802.3ab task force   ꢀ UTP   ꢂ Uses all 4 line pairs simultaneously for duplex   transmission! (echo cancellation)   ꢂ 5 level PAM coding   • 4 levels encode 2 bits + extra level used for Forward   Error Correction (FEC)   ꢂ Signal rate: 4 x 125 Mbaud = 4 x 250Mbit/s data   rate   • Cat. 5 links, max 100 m; all 4pairs, cable must   conform to the requirements of ANSI/TIA/EIA-568-A   ꢂ Only 1 CSMA/CD repeater allowed in a   collision domain   18   (C) Herbert Haas 2005/03/11   It is very difficult to transmit Gigabit speeds over unshielded twisted pair cables.   Only a mix of multiple transmission techniques ensure that this high data rate can   be transmitted over a UTP Cat5 cable. For example all 4 pairs are used together   for both directions. Echo cancellation ensures that the sending signal does not   confuse the received signal. 5 level PAM is used for encoding instead of 8B10B   because of its much lower symbol rate. Now we have only 125 Mbaud x 4   instead of 1250 Mbaud.   The interface design is very complicated and therefore relatively expensive.   Using Cat 6 or Cat 7 cables allow 500 Mbaud x 2 pairs, that is 2 pairs are   designated for TX and the other 2 pairs are used for RX. This dramatically   reduces the price but requires better cables, which are not really expensive but   slightly thicker. Legacy cable ducts might be too small in diameter.   18   Download from Www.Somanuals.com. All Manuals Search And Download.   Several Physical Media Supported   Logical Link Control LLC   MAC Control (optional)   Data Link Layer   Media Access Control MAC   PLS   AUI   Reconciliation   Reconciliation   Reconciliation   GMII   MII   PLS   AUI   MII   PCS   PMA   PMD   MDI   PCS   PHY   PMA   PMA (MAU)   MDI   PMA   MDI   PMD   MDI   Medium   Medium   Medium   Medium   1-10 Mbit/s   10 Mbit/s   100 Mbit/s   1000 Mbit/s   AUI Attachment Unit Interface, PLS   PCS Physical Coding Sublayer, MII   Physical Layer Signaling, MDI   Medium Dependent Interface   Gigabit Media Independent   Media Independent Interface, GMII   Interface, PMA Physical Medium Attachment, MAU Medium Attachment Unit, PMD Physical Medium   Dependent   19   (C) Herbert Haas 2005/03/11   The diagram above shows various physical media designs supported by the   official GE standard. Each modern GE card could theoretically support the old   10 Mbit/s standard as well. However many vendors create GE NICs that only   support GE or GE and FE—who would connect a precious GE interface with   another interface, which is 100 times slower?   19   Download from Www.Somanuals.com. All Manuals Search And Download.   10 Gigabit Ethernet / IEEE 802.3ae   ꢀ Only optical support   ꢂ 850nm (MM) / 1310nm /1550 nm (SM only)   ꢂ No copper PHY anymore !   ꢀ Different implementations at the   moment – standardization not finished!   ꢀ 8B/10B (IBM), SONET/SDH support, …   ꢀ XAUI ("Zowie") instead of GMII   20   (C) Herbert Haas 2005/03/11   10 GE only supports optical links. Note that GE is actually a synchronous   protocol! There is no statistical multiplexing done at the physical layer anymore,   because optical switching at that bit rate only allows synchronous transmissions.   The GMII has been replaced (or enhanced) by the so-called XAUI, known as   "Zowie".   Note: At the time of writing this module, the 10 GE standard was not fully   finished. Though, some vendors already offer 10 GE interface cards for their   switches.   These interfaces are very expensive but the investment ensures backward   compatibility to lower Ethernet rates and at the same time provides a very high   speed WAN interface.   An alternative technology would be OC192, which requires a very expensive and   complex SONET/SDH environment.   20   Download from Www.Somanuals.com. All Manuals Search And Download.   Note   ꢀ GE and 10GE use synchronous   physical sublayer !!!   ꢀ Recommendation: Don't use GE over   copper wires   ꢂ Radiation/EMI   ꢂ Grounding problems   ꢂ High BER   ꢂ Thick cable bundles (especially Cat-7)   21   (C) Herbert Haas 2005/03/11   Both GE and 10GE are synchronous physical technologies on fiber. It not   recommended to use GE over copper wires anymore although 802.3ab would   specify it. This is because the whole electrical hardware (cables and connectors)   are re-used from older Ethernet technologies and have not been designed to   support such high frequencies.   For example the RJ45 connector is not HF proof. Furthermore, shielded twisted   pair cables require a very good grounding, seldom found in reality. The Bit Error   Rate (BER) is typically so high that the effective data rate is much lower than   GE, for example 30% only.   21   Download from Www.Somanuals.com. All Manuals Search And Download.   Summary   ꢀ Ethernet evolved in the opposite direction:   ꢂ Collision free   ꢂ WAN qualified   ꢂ Switched   ꢀ Several coding styles ꢁ Complex PHY   architecture   ꢀ Plug & play through autonegotiation   ꢀ Much simpler than ATM but no BISDN   solution – might change!   22   (C) Herbert Haas 2005/03/11   22   Download from Www.Somanuals.com. All Manuals Search And Download.   Quizz   ꢀ Why tends high-speed Ethernet to   synchronous PHY?   ꢀ Can I attach a 100 Mbit/s port to a   1000 Mbit/s port via fiber?   ꢀ What is the idea of Etherchannels?   (Maximum bit rate, difference to   multiple parallel links)   23   (C) Herbert Haas 2005/03/11   23   Download from Www.Somanuals.com. All Manuals Search And Download.   Hints   ꢀ Q1: On fiber its difficult to deal with   asynchronous transmission, photons   cannot be buffered easily, store and   forward problems   ꢀ Q2: No, autonegotiation on fiber   does not care for data rates   ꢀ Q3: "normal" parallel links would be   disabled by STP, Etherchannel   supports up to 8 links   24   (C) Herbert Haas 2005/03/11   24   Download from Www.Somanuals.com. All Manuals Search And Download.   |